Optimized pulsed write schemes improve linearity and write speed for low-power organic neuromorphic devices

Scott T. Keene, Armantas Melianas, Elliot J. Fuller, Yoeri van de Burgt, A. Alec Talin, Alberto Salleo

Research output: Contribution to journalArticleAcademicpeer-review

9 Citations (Scopus)
1 Downloads (Pure)

Abstract

Neuromorphic devices are becoming increasingly appealing as efficient emulators of neural networks used to model real world problems. However, no hardware to date has demonstrated the necessary high accuracy and energy efficiency gain over CMOS in both (1) training via backpropagation and (2) in read via vector matrix multiplication. Such shortcomings are due to device non-idealities, particularly asymmetric conductance tuning in response to uniform voltage pulse inputs. Here, by formulating a general circuit model for capacitive ion-exchange neuromorphic devices, we show that asymmetric nonlinearity in organic electrochemical neuromorphic devices (ENODes) can be suppressed by an appropriately chosen write scheme. Simulations based upon our model suggest that a nonlinear write-selector could reduce the switching voltage and energy, enabling analog tuning via a continuous set of resistance states (100 states) with extremely low switching energy (∼170 fJ • μm-2). This work clarifies the pathway to neural algorithm accelerators capable of parallelism during both read and write operations.

Original languageEnglish
Article number224002
Number of pages8
JournalJournal of Physics D: Applied Physics
Volume51
Issue number22
DOIs
Publication statusPublished - 8 May 2018

Fingerprint

linearity
Tuning
tuning
Electric potential
Backpropagation
selectors
Particle accelerators
Energy efficiency
Ion exchange
electric potential
multiplication
Neural networks
Hardware
energy
CMOS
hardware
accelerators
education
Networks (circuits)
nonlinearity

Keywords

  • electrochemical organic neuromorphic device
  • neural network
  • neuromorphic computing
  • organic electronics
  • PEDOT:PSS
  • resistive memory
  • symmetric cycling

Cite this

Keene, Scott T. ; Melianas, Armantas ; Fuller, Elliot J. ; van de Burgt, Yoeri ; Talin, A. Alec ; Salleo, Alberto. / Optimized pulsed write schemes improve linearity and write speed for low-power organic neuromorphic devices. In: Journal of Physics D: Applied Physics. 2018 ; Vol. 51, No. 22.
@article{a3d730c197ad4a8996728bc80086a0bd,
title = "Optimized pulsed write schemes improve linearity and write speed for low-power organic neuromorphic devices",
abstract = "Neuromorphic devices are becoming increasingly appealing as efficient emulators of neural networks used to model real world problems. However, no hardware to date has demonstrated the necessary high accuracy and energy efficiency gain over CMOS in both (1) training via backpropagation and (2) in read via vector matrix multiplication. Such shortcomings are due to device non-idealities, particularly asymmetric conductance tuning in response to uniform voltage pulse inputs. Here, by formulating a general circuit model for capacitive ion-exchange neuromorphic devices, we show that asymmetric nonlinearity in organic electrochemical neuromorphic devices (ENODes) can be suppressed by an appropriately chosen write scheme. Simulations based upon our model suggest that a nonlinear write-selector could reduce the switching voltage and energy, enabling analog tuning via a continuous set of resistance states (100 states) with extremely low switching energy (∼170 fJ • μm-2). This work clarifies the pathway to neural algorithm accelerators capable of parallelism during both read and write operations.",
keywords = "electrochemical organic neuromorphic device, neural network, neuromorphic computing, organic electronics, PEDOT:PSS, resistive memory, symmetric cycling",
author = "Keene, {Scott T.} and Armantas Melianas and Fuller, {Elliot J.} and {van de Burgt}, Yoeri and Talin, {A. Alec} and Alberto Salleo",
year = "2018",
month = "5",
day = "8",
doi = "10.1088/1361-6463/aabe70",
language = "English",
volume = "51",
journal = "Journal of Physics D: Applied Physics",
issn = "0022-3727",
publisher = "Institute of Physics",
number = "22",

}

Optimized pulsed write schemes improve linearity and write speed for low-power organic neuromorphic devices. / Keene, Scott T.; Melianas, Armantas; Fuller, Elliot J.; van de Burgt, Yoeri; Talin, A. Alec; Salleo, Alberto.

In: Journal of Physics D: Applied Physics, Vol. 51, No. 22, 224002, 08.05.2018.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Optimized pulsed write schemes improve linearity and write speed for low-power organic neuromorphic devices

AU - Keene, Scott T.

AU - Melianas, Armantas

AU - Fuller, Elliot J.

AU - van de Burgt, Yoeri

AU - Talin, A. Alec

AU - Salleo, Alberto

PY - 2018/5/8

Y1 - 2018/5/8

N2 - Neuromorphic devices are becoming increasingly appealing as efficient emulators of neural networks used to model real world problems. However, no hardware to date has demonstrated the necessary high accuracy and energy efficiency gain over CMOS in both (1) training via backpropagation and (2) in read via vector matrix multiplication. Such shortcomings are due to device non-idealities, particularly asymmetric conductance tuning in response to uniform voltage pulse inputs. Here, by formulating a general circuit model for capacitive ion-exchange neuromorphic devices, we show that asymmetric nonlinearity in organic electrochemical neuromorphic devices (ENODes) can be suppressed by an appropriately chosen write scheme. Simulations based upon our model suggest that a nonlinear write-selector could reduce the switching voltage and energy, enabling analog tuning via a continuous set of resistance states (100 states) with extremely low switching energy (∼170 fJ • μm-2). This work clarifies the pathway to neural algorithm accelerators capable of parallelism during both read and write operations.

AB - Neuromorphic devices are becoming increasingly appealing as efficient emulators of neural networks used to model real world problems. However, no hardware to date has demonstrated the necessary high accuracy and energy efficiency gain over CMOS in both (1) training via backpropagation and (2) in read via vector matrix multiplication. Such shortcomings are due to device non-idealities, particularly asymmetric conductance tuning in response to uniform voltage pulse inputs. Here, by formulating a general circuit model for capacitive ion-exchange neuromorphic devices, we show that asymmetric nonlinearity in organic electrochemical neuromorphic devices (ENODes) can be suppressed by an appropriately chosen write scheme. Simulations based upon our model suggest that a nonlinear write-selector could reduce the switching voltage and energy, enabling analog tuning via a continuous set of resistance states (100 states) with extremely low switching energy (∼170 fJ • μm-2). This work clarifies the pathway to neural algorithm accelerators capable of parallelism during both read and write operations.

KW - electrochemical organic neuromorphic device

KW - neural network

KW - neuromorphic computing

KW - organic electronics

KW - PEDOT:PSS

KW - resistive memory

KW - symmetric cycling

UR - http://www.scopus.com/inward/record.url?scp=85047206903&partnerID=8YFLogxK

U2 - 10.1088/1361-6463/aabe70

DO - 10.1088/1361-6463/aabe70

M3 - Article

VL - 51

JO - Journal of Physics D: Applied Physics

JF - Journal of Physics D: Applied Physics

SN - 0022-3727

IS - 22

M1 - 224002

ER -